Contents lists available at ScienceDirect

Habitat International

journal homepage: www.elsevier.com/locate/habitatint

GIS coupled multiple criteria decision making approach for classifying urbancoastal areas in India

Ravinder Dhimana, Pradip Kalbara,b,∗, Arun B. Inamdarc

a Centre for Urban Science and Engineering (CUSE), Indian Institute of Technology Bombay, Powai, Mumbai 400076, Indiab Interdisciplinary Programme in Climate Studies, Indian Institute of Technology Bombay, Powai, Mumbai 400076, Indiac Centre of Studies in Resources Engineering (CSRE), Indian Institute of Technology Bombay, Powai, Mumbai 400076, India

A R T I C L E I N F O

Keywords:Coastal citiesCoastal regulation zoneUrbanizationMultiple criteria decision makingGISSustainable development

A B S T R A C T

Coastal area classification in India is a challenge for decision makers due to unclear directions in implementationof coastal regulations and lack of scientific rational about existing classification methods. To improve the ob-jectivity of the coastal area classification is the aim of the present work. A Geographical Information System(GIS) coupled Multi-criteria Decision Making (MCDM) approach is developed in this work to provide scientificrational for classifying coastal areas. Utility functions are used to transform the physical coastal features intoquantitative membership values. Different weighting schemes for coastal features are applied to derive CoastalArea Index (CAI) which classifies the coastal areas in distinct categories. Mumbai, the coastal megacity of India,is used as case study for demonstration of proposed approach. Results of application of GIS-MCDM approachshowed the clear demarcation of coastal areas based on CAI is possible which provides a better decision supportfor developmental and planning authorities to classify coastal areas. Finally, uncertainty analysis using MonteCarlo approach to validate the sensitivity of CAI under different scenarios was carried out.

1. Introduction

India, surrounded by Arabian Sea, Bay of Bengal and Indian Oceanwith a 7500 km long coastline is inhabited by nearly 25% of populationwithin the vicinity of 50 km towards the coast (Nayak, Chandramohan,& Desai, 1992). Indian coastal states support national economy withdevelopmental activities (largest in the South Asian Region) and nu-merous marine resources located in 2.02 million km2 Exclusive Eco-nomic Zone (EEZ). At another hand, these states are under extremepressure of natural hazards because of vulnerable geographical loca-tions (Alcántara-Ayala, 2002; Smith & Petley, 2008) and anthropogenicdevelopment (Nair & Gopinathan, 1981; Ramesh, Lakshmi, George, &Purvaja, 2015). India harbors some of the world's fastest growingcoastal megacities such as Mumbai and Kolkata, with Chennai joiningthe league shortly.

The implications of coastal cities are pollution and natural hazardswhich are inevitable and the effect can be destructive in urban coastalregions with huge population (Cenci et al., 2015; Nayak et al., 1992;Newton, Carruthers, & Icely, 2012). Specifically (Vaz, 2014), reportsthat in Mumbai there is observed transformation of 36% of mangroveeco-system area to urban land since 1973 due to urbanization. Along

with urbanization there is industrial belt development along the Indiancoasts causing effluent discharges from cities and industrial belts(Datta, Chakraborty, Jaiswar, & Ziauddin, 2010; Ramachandran, 1999;Shirodkar et al., 2009). On the east coast, increased metal concentra-tion in the coastal waters of Pondicherry as reported by Govindasamyand Azariah (1999) was a result of the effluent discharges of nearly 16major and minor industries.

1.1. Coastal regulations in India

Coastal area classification for developmental activities in urbanareas is a challenging task due to conflicts among associated stake-holders. Different countries have various mechanism to manage theircoastal resources with the help of policies. These policies are aimed atcreating balance between human development and coastal biodiversity.Similar efforts have been made by State and Central Government inIndia to direct the development in the coastal areas using regulatoryframeworks. One of the most prevailing regulation is CoastalRegulation Zone (CRZ) policy (Krishnamurthy, DasGupta, Chatterjee, &Shaw, 2014). The Ministry of Environment, Forest and Climate Change(MoEF & CC) had issued the Coastal Regulation Zone (CRZ)

https://doi.org/10.1016/j.habitatint.2017.12.002Received 21 September 2017; Received in revised form 10 November 2017; Accepted 1 December 2017

∗ Corresponding author. Centre for Urban Science and Engineering (CUSE), Associated Faculty – Interdisciplinary programme (IDP) in Climate Studies, Indian Institute of TechnologyBombay, Powai, Mumbai 400 076, India.

E-mail address: kalbar@iitb.ac.in (P. Kalbar).

Habitat International 71 (2018) 125–134

0197-3975/ © 2017 Elsevier Ltd. All rights reserved.

T

Notification, 1991 under the Environment (Protection) Act, 1986 andthere have been continuous amendments in this notification since then.The CRZ notification was issued with a vision of sustainable develop-ment and conservation of marine resources.

Present coastal regulations (CRZ, 2011) has issues related to im-plementation at administrative level and violations at private/govern-ment level. Considerable variations have been observed in the appli-cation and interpretation of coastal regulations in various Indian states(Sonak, Pangam, & Giriyan, 2008). Lack of clarity and definitions inCRZ regulations has led to subjective interpretations. This subjectivitystems from the basic definitions used to classify the coastal areas. Forexample, in prevailing CRZ regulation, areas come in CRZ-I are ecolo-gically subtle areas like mangroves (if area more than 1000 squaremeters, a buffer area of 50 m shall be provided), corals, sand dunes, saltmarshes, bird & turtle nesting grounds, horse shoe crab habitats, seagrass beds, mudflats and the area between Low Tide Line and High TideLine and it's hard to demarcate these areas independently. Chouhan,Parthasarathy, and Pattanaik (2016) reported the violation of coastalregulations in Mumbai city leading to the destruction of ecosystem andbiodiversity in coastal region due to rapid uneven population growthand developmental activities. Parthasarathy (2011) discussed the landencroachment issues and its associated effects on coastal communitiesin Mumbai. Ignorance of coastal communities existence by adminis-trations has resulted into commercial exploitation and loss of biodi-versity despite the coastal communities are meant to be protected ac-cording to CRZ (2011) regulations (Krishnamurthy et al., 2014).Absolute gap at institutional level for CRZ implementation has beenreported by academicians and expert's working in the area of coastalresource management in India (R Dhiman, June 2015, Personal Com-munication).

There is a need to bring all the priorities related to the development,natural disasters and environment in decision-making related withIndian coastal cities (Srinivas & Nakagawa, 2008). All these issues re-lated to futile regulations, policy violations and fragile institutionalframework urge the requirement for management of coastal resourcesin a sustainable manner which is possible through suitable planning andpolicy improvement. Quantitative decision analysis is one of the sui-table approaches answering strategic or policy decisions with largeuncertainties and multiple conflicts (Zhou, Ang, & Poh, 2006). There-fore, present work focuses on improvement for coastal area classifica-tion methods in Indian cities through advancement in decision analysis.

1.2. Use of GIS and MCDM for area classification

Geographical Information System (GIS) tool is a set of computerhardware and software which is used to process and store the spatiallyreferenced data to derive information (Burrough & McDonnell, 1998).One of the key application of GIS is land suitability analysis (Collins,Steiner, & Rushman, 2001) targeting the identification of appropriatespatial configurations for upcoming land uses based on specific criteriaand requirements (Brail & Klosterman, 2001). Specific land suitabilityapplications of GIS include urban and regional planning (Baud, Scott,Pfeffer, Sydenstricker-Neto, & Denis, 2015; Janssen & Rietveld, 1990),selecting site for public and private sector facilities (Longley,Goodchild, Maguire, & Rhind, 2005), habitant selection for animal andplant species (Schmoldt, Kangas, Mendoza, & Pesonen, 2001), en-vironmental impact assessment (Gontier, Mörtberg, & Balfors, 2010),coastal vulnerability assessment (Le Cozannet et al., 2013; Mani Murali,Ankita, Amrita, & Vethamony, 2013).

GIS is acknowledged as an appropriate decision support tool foranalyzing and solving problems with integration of spatially referenceddata (Brown & Kyttä, 2014). Major limitation of GIS to be designate ascomplete decision support tool is that, GIS doesn't incorporate pre-ferences (evaluation criteria/alternatives) of decision makers(Malczewsk, 1999). Coupling GIS with Multiple Criteria DecisionMaking (MCDM) approach provides better decision support and has

been widely practiced (Malczewski, 2006). A detailed literature surveyby Keefer, Kirkwood, and Corner (2004) has highlighted the significantgrowth in use of decision analysis approach for decision making ininterdisciplinary areas. One of the most suitable fundamental decisionanalysis approach is MCDM (Feizizadeh & Blaschke, 2014). Coupling ofGIS with MCDM can transform the geographical data and combine thepreferences from decision makers (value judgments) to derive the in-formation for decision making (Drobne & Lisec, 2009). Capabilities ofGIS coupled MCDM methods can be utilized for benefits of advance-ment in theoretical and applied research areas (Malczewski, 2004).

In literature, we observed numerous applications of coupling GISwith MCDM in the area of agricultural resource management (Ananda& Herath, 2003a), natural resource management (Mendoza & Martins,2006), urban planning (Bryan, 1988), forest management (Ananda &Herath, 2003b), environmental management (Howard, 1991; Munda,Nijkamp, & Rietveld, 1993), energy policy analysis (Greening &Bernow, 2004), wetland management (G A; Mendoza & Martins, 2006),selection of storm water management options (Gogate, Kalbar, & Raval,2017) and food security (Renwick, 2015). Alves, Coelho, Coelho, andPinto (2011) interestingly demonstrates development of coastal zonerisk map with the help of vulnerability modelling for the northwestPortuguese Coastal Zone.

Some of literature studies have attempted Analytical HierarchicalProcess (AHP) methods for development of disaster resilient index toassess coastal communities (Orencio & Fujii, 2013), conservation ofcoastal areas in Iran (Pourebrahim, Hadipour, Mokhtar, & Taghavi,2014), comprehensive framework for coastal reclamation in China(Feng, Zhu, & Sun, 2014), optimizing strategies for protection of coastallandscape (Baby, 2013), protection priority in coastal environment ofTaiwan (Chang, Liou, & Chen, 2012) and coastal vulnerability assess-ment (Sudha Rani, Satyanarayana, & Bhaskaran, 2015). However, uti-lity based and compensatory approaches have not been attempted so farfor management of coastal areas in Indian coastal cities.

1.3. Problem definition

The above literature suggests that there is lack of quantitative sci-entific rational method applied presently in India for classifying coastalareas. The present work attempts to fulfill this gap by developing acoastal area classification method based on GIS coupled MCDM ap-proach. Main objective of this research is to develop area classificationmethod which is technically robust and easy to use for decision makingin coastal planning. Such a decision-making method will synergies de-velopmental activities and environment conservation in coastal cities ofIndia by appropriately classifying coastal areas using quantitative ap-proach.

2. Methods

Present study relies on application of GIS and MCDM for coastalareas. A case study is used to demonstrate the application of themethod. Results obtained from the case study are validated and rig-orous sensitivity analysis was performed. The overall methodologyfollowed in this work is represented in Fig. 1. Next sub-sections de-scribes the case study and the methods in details.

2.1. Study area

We selected Mumbai, Indian coastal megacity as primary study areafor validation of coastal area classification method using GIS coupledMCDM approach. Coastal area of Mumbai has high environmentalvalue territory exposed to anthropic activities. Furthermore, this coastalarea is subjected to major urban issues related to water shortage, sea-sonal monsoonal flooding, untreated pollution discharge in adjacentcreeks, direct solid waste dumping in mangroves (Pacione, 2006). Se-vere coastal pollution leading to depletion of coastal ecosystem made

R. Dhiman et al. Habitat International 71 (2018) 125–134

the coastal areas unsuitable for environment management and sus-tainable planning of developmental activities (Murthy, Rao, & Inamdar,2001).

Present study derives primary datasets from remote sensing which isfurther processed in GIS and secondary datasets from literature, andother national as well as state level government agencies. A detaileddescription about sources of data is given in Table 1.

2.2. Data processing

This study used remote sensing dataset for year 2016 from

LANDSAT 8 (USGS) with a resolution of 30 m for different coastalfeatures (refer to Table 1) for Mumbai coast where we have consideredthe area within the 1 km range towards land from high tide at intactshoreline of Mumbai coast. Dataset in GIS was processed for estimationof Land Use, Land Cover (LULC) and coastal geomorphology for year of2016 using supervised image classification technique (Figure S1 inSupplementary Information 1). Land Use in the area is dominated byurbanization (habitation/built-up land) and developmental activities(salt pan, ports) whereas Land Cover is dominated by mangroves,mudflats and marshy land. Some part of the study area is covered undersandy beaches, rocks and vegetation. Coastal slope and elevationmeasurements are evaluated using satellite imagery from SRTM datasetin GIS (Figure S2, S3). Coastal slope of study area is< 25% and most ofthe area is below 10 m of elevation characterizing the flat or low-lyingtopography. Soil and Geology maps for Mumbai city are used as sec-ondary data from Survey of India as shown in Figure S4 and S5 re-spectively. Raster images of Soil and Geology maps are digitized andgeo-referenced in GIS environment. In Mumbai, most land is reclaimedusing backfilled soil during the course of the city's development. Studyarea has majority of reclaimed and marshy soil so there are three majorclasses of soil type - mud, backfilled and clay. Normalized DifferenceVegetation Index (NDVI) maps are being used in agriculture, forestry,ecology and more. Healthy vegetation (or chlorophyll) reflects morenear-infrared (NIR) and green light compared to other wavelengths. It

Fig. 1. Methodology for computation of CAI using GIS coupled MCDM approach.

Table 1Types and Sources of Dataset used for GIS coupled MCDM approach.

Coastal Features Source Resolution (meters)

Geomorphology Landsat 8 30Land Use Landsat 8 30Land Cover Landsat 8 30Shorelines/Inter Tidal Zone Landsat 8 30Elevation SRTM 30NDVI Landsat 8 30Coastal Slope SRTM 30Geology Survey of India Vector DatasetSoil Survey of India Vector Dataset

R. Dhiman et al. Habitat International 71 (2018) 125–134

127

absorbs more red and blue light. We have derived NDVI map fromLandsat 8 satellite imagery where calculations of NDVI for a given pixelalways result in a number that ranges from minus one (−1) to plus one(+1); however, no green leaves gives a value close to zero (Figure S6).Negative values (0 to −1) represent water and zero means no vegeta-tion and close to +1 (0.8–0.9) indicates the highest possible density ofgreen leaves.

Another important aspects for coastal planning is to understand theprocess of erosion, accretion, sediment transportation along theshoreline to understand the regional dynamic behaviors of the coast(Murali et al., 2015). In recent study, by Cenci et al. (2017) integrationof remote sensing and GIS techniques were attempted for managementof crucial coastal areas with the help of monitoring and modelling ofshoreline changes. We assessed the shoreline change and Inter TidalZone (gives insights about erosion and accretion happening at parti-cular location) at temporal scale (2003–2017) using data from variousLandsat Missions which is presented in Figure S7. Digital ShorelineAnalysis System (DSAS) in GIS is a useful tool for mapping shorelinechange and the same is used for this study. Intertidal Zone is the areabetween HTL (High Tide Line) and LTL (Low Tide Line) and this area ismeasured using the difference between temporal shorelines at decadalscale in the current study. Further, entire dataset was transformed intoraster format which produced pixel based layers for each availablecoastal feature.

2.3. GIS coupled MCDM approach

In this work a novel method for classification of coastal areas basedon coastal features in GIS environment using MCDM approach (Fig. 2)is introduced. MCDM, a type of decision analysis, is a structured fra-mework for analyzing decision problems characterized by complexmultiple objectives. MCDM can also deal with long-term time horizons,uncertainties, risks and complex value issues (Nijkamp, Rietveld, &Voogd, 1990). Ananda and Herath (2009) has suggested interactive useof the MCDM methods to improve the efficiency of the planning processand advocated the use of more than just one MCDM method or a hybridapproach. GIS-based multi-criteria decision making (GIS-MCDM) ap-proach has extended in recent times to solve planning problems thatinvolve conflicting multi-objectives such as land use allocation pro-blems (Eastman, 1995; Janssen & Rietveld, 1990; Yeh & Li, 1998). Inthis work Coastal Area Index (CAI) based on GIS-MCDM approach isdeveloped by applying utility membership functions to coastal featuresfor classification of coastal areas according to their importance andsensitivity towards developmental activities.

2.3.1. Use of utility functions for area classificationA utility function can be defined as a mathematical function of

preferences assigned to goods or services. Typically, utility functioncaptures the stakeholders’ preferences by assigning a utility number toeach group of goods and services. Some of the applications of utilityfunction in interdisciplinary research are related to protection ofmarine area (Fernandes, Ridgley, & Van’t Hof, 1999), water policy andsupply planning (Joubert, Stewart, & Eberhard, 2003), ground watermanagement (Almasri & Kaluarachchi, 2005), measure of welfare in-dices (Hajkowicz, 2006) available in literature. Multiple criteria ana-lysis problems deals with indicators having different units which re-quire transformation using a common scale using utility function,mostly 0 to 1, so that indicators can be meaningfully integrated in theoverall index (or score).

There are total 6 coastal features i.e. layers (LULC, Geology, Soil,NDVI, Slope and DEM) used for deriving CAI in this work. We estab-lished a distinct membership function for each of this layers, trans-forming individual pixel values into membership values. The estab-lished membership function is shown in Fig. 2 for LULC layer and forother coastal features the defined membership functions are given inFigure S8 (a - e). The membership values obtained by applying mem-bership functions to coastal features are used for computing CAI.

For example in Fig. 2, for LULC layer, highest importance or max-imum utility value (0.9 and 1) is given to mangroves and water bodieslike creeks and estuaries respectively. These higher utility values aresupported by the fact that mangroves plays important role in preservingcoastal ecosystem. Newton et al. (2012) and Rog, Clarke, and Cook(2017) gives detailed review about importance of mangroves in ter-restrial ecosystem. Following the same analogy of valuing the eco-system services aquatic vegetation was given a membership value of0.9. Marshes, forest and mudflats were assigned a membership value of0.8 based on their importance in primary productivity of coastal system(Burford et al., 2016). Similarly, according to the comparative study onfaunal biodiversity of Mumbai Coast (Datta et al., 2010), a membershipvalue of 0.6 and 0.5 was assigned to vegetation and crop lands andsandy areas respectively. Referring to literature providing importanceof coastal feature such salt pan, fallow land a membership value of 0.4and 0.3 was assigned respectively (Abadie, Galarraga, & de Murieta,2017; Surjan, Parvin, Atta-ur-Rahman, & Shaw, 2016).

2.3.2. Development of Coastal Area Index (CAI)We computed the CAI using membership function values at pixel

level for coastal features against individual layer of available dataset. Adetailed framework for method is shown in Fig. 1 and computationprocedure for CAI is presented in Fig. 3. The membership functions foreach layers are derived in normalized scales (0–1) using the earliermentioned approach. According to importance of these layers ad-hocweights are assigned based on authors’ understanding of the specificcase study of Mumbai (refer to Table 2). Further, the normalized valuesof pixels are multiplied with weights of the respective layers and ag-gregated by Linear Weighted Sum (LWS) method to obtain CAI.

CAI for the ith pixel can be computed using following equation:

∑= = …=

CAI w x for i m. 1,2,3 ..,ij

n

j ij1 (1)

Where,

i is pixel number in the layer (there are m number of pixel in the onelayer)j is layer number (there are n number of layers)xij is the pixel value of the ith pixel in the jth layerwj is the weight of the jth layer

The CAI obtained using Eq. (1) is further normalized to express on ascale of 0–10. To check the sensitivity of the applied weighting schemeFig. 2. Utility function applied for Land Use and Land Cover.

R. Dhiman et al. Habitat International 71 (2018) 125–134

128

two different weights sets are also tested as shown in Table 2 and ex-plained in next section.

Classification of coastal areas in CAI is based on the sensitivity ofcoastal features towards developmental activities where Class 1 is ex-tremely sensitive, Class 2 is highly sensitive, Class 3 is moderatelysensitive and Class 4 is low sensitive against the anthropogenic devel-opment. In CAI, Class 1 is the intertidal zone of coastal Mumbai whichis highly dynamic in nature and much important for demarcation ofland from sea. Intertidal zones contain high nutrients which supportsthe productivity of marine life and hence this is considered as no de-velopment zone for any kind of developmental activities. Class 2 re-present the highly sensitive areas including mangrove covers andmudflats along the coastline and any kind of development should beavoided except environmental conservational activities. Class 3 ismoderately sensitive zone and shows the balanced areas for develop-mental and conservational activities where human interventions can beallowed after comprehensive environmental impact assessment anddeveloping mitigation plans. Finally, Class 4 in CAI is low sensitivebecause this area is already overexploited for development resultinginto depletion of sensitive environmental features. The proposedmethod for CAI clearly demarcated the boundary lines among these 4classified areas which is useful for planning process by associated sta-keholders such as planning authorities and Urban Local Bodies (ULBs).

2.3.3. Sensitivity analysisRigorous sensitivity analysis has been carried out to assess the un-

certainty and sensitivity of weights assigned to the layers for calculationof CAI using Monte Carlo simulation. Three different scenarios areconsidered, ad-hoc weights (given by authors considering under-standing about case study), equal weights and random weights forlayers (coastal features). In this setup, 1000 random numbers withvariation of± 50 of each scenario listed in Table 2 were generated.These values were further transformed and normalized to get theweighting norm as 1 for each of the iteration. The difference between

the CAI values obtained using given set of weights and the mean of theCAI results of 1000 test runs normalized with standard deviation, whichis basically a z-score, serves as a good indicator of the reliability of theweights (Benke & Pelizaro, 2010; Feizizadeh & Blaschke, 2014).

3. Results and discussions

In our present study, we assessed the major coastal features re-presenting physical characteristics viz., – LULC, DEM, Slope, NDVI, Soiland Geology of Mumbai coast. These feature are further quantified intosub-classes at 30 m spatial resolution. The detail maps of these sub-classes are provided in Figure S1-S6 in SI. Mumbai coast is observedunder permanent erosion due to developmental activities at specificlocations whereas accretion and erosions at seasonal scale are observedalong the entire coastline (Figure S7). This assessment of coastal fea-tures provides deeper understanding of spatial distribution of physicalcharacteristics along Mumbai coast.

3.1. Results of CAI

The main result of this study is CAI for case study area which isshown at 30 m resolution in Fig. 4. As evident from Fig. 4, the CAIdemarcates coastal areas in different classes of sensitiveness, which isbasically prioritization of coastal areas. Extremely sensitive area isdenoted by class 1 which is intertidal zone and it is extremely dynamic(daily and seasonal scale) in nature. Class 1 is independent of compu-tation procedure of CAI as the area which comes under intertidal zone isdirectly considered as extremely sensitive. Class 2 is designated ashighly sensitive and CAI index range for this class is 6–10, majorlydistributed along the Thane Creek in Mumbai considering the biodi-versity including mangroves, mudflats and other important indicatorsof primary productivity in the area. Area classified under class 3 of CAIwithin a range of 3–6 is showing moderate sensitivity because this areais majorly influenced by open spaces and barren land which can beconsidered for both conservational and developmental activities as perthe requirement in the area. Low sensitive area in CAI is represented byclass 4 with index range of 0–3, depicting most of the urbanized area inMumbai City.

To validate the derived CAI in this work, a small area of540 m × 540 m in Colaba area in Mumbai was selected. As shown inFig. 5, each layer representing individual coastal feature is illustrated atpixel level (30 m). For example, for LULC layer in this small area of

Fig. 3. Model description for computing CoastalArea Index in GIS.

Table 2Different Scenarios of weights for sensitivity analysis of CAI.

LULC DEM GEO NDVI SLOPE SOIL

Ad-hoc Weights 30 20 10 15 15 10Random Weights 10 10 25 30 20 05Equal Weights 16.6 16.6 16.6 16.6 16.6 16.6

R. Dhiman et al. Habitat International 71 (2018) 125–134

129

0.29 km2, four distinct sub-classes are present which are transformedinto membership values using membership function given in Fig. 2. Insimilar way other coastal features are also transformed into respectivemembership values. Layer weights are also shown in Fig. 5. Using themathematical approach shown in Fig. 3, the CAI for each of this pixel isobtained and shown in Fig. 5. If we compare the obtained CAI value foreach pixel with satellite imagery of the Colaba region under evaluation,the CAI values objectively classifies this area into distinct CAI indices.For example, we have demonstrated the calculation of CAI for onecorner most (southernmost) pixel in Fig. 5. For this pixel the CAI wasestimated to be 6 and the area attributed for this pixel in the image isvegetation surrounded by built up area, which was also confirmed bythe field visit to this area, hence it is classified as moderately sensitive.This shows that CAI is a robust measure that can be used for classifyingurban coastal areas.

3.2. Sensitivity analysis

The CAI in this study is derived using applying weights to the layers.This induces uncertainty in the CAI values. CAI results for first scenario(based on ad-hoc weights) are shown in Fig. 4. For the other two sce-narios results of CAI are shown in Figure S9 (a, b). These results of CAIsshows significant disagreement between the three scenarios.

In Monte Carlo simulation setup for sensitivity analysis, we obtainedthe z-score which is the difference between the CAI results of a givenad-hoc weight and the mean of 1000 test runs of CAI values normalizedusing standard deviation obtained using ad-hoc weights. Results of thistest are shown in Fig. 6 (a). The z-score lies between −1 and 1 for ad-hoc weight scenario.

Similar approach of quantifying sensitivity for equal weights andrandom weights are followed and results are presented in Fig. 6 (b, c).These figures shows that high sensitivity of the applied weightingschemes in both scenarios. For example, the difference between meanof the CAI results of 1000 test runs and the CAI values obtained using

Fig. 4. Coastal Area Index for Mumbai showing sensitivitybased classification of coastal areas.

R. Dhiman et al. Habitat International 71 (2018) 125–134

130

equal weights, the z-score of this test, lies in the range of −1.5 to 1.5whereas for random weights z-score lies in the range of −2 to 2. Thesevalues are almost 2 times higher than the ad-hoc weighting schemescenario results of sensitivity analysis. Hence, it is shown from theseresults that, ad-hoc weights are more robust in providing CAI for givencase study (i.e. ad-hoc weights are less uncertain while compared withequal and random weights). Further, it can be argued that in com-parative assessment of three scenarios with variation in layer weightsshows that ad-hoc weights obtained using domain experts are mostdecisive in uncertain conditions.

To ascertain the sensitivity results towards applied aggregationmethod for deriving CAI, we also checked the correlations among dif-ferent layers of coastal features and results are presented in Table 3.There exists no significant correlation among the layers which suggestthat all different layers are independent and therefore there is no in-terdependence between these layers. Hence, as recently reported in(Kalbar, Birkved, Nygaard, & Hauschild, 2016) if the indicators are notinterdependent then simple aggregation methods such as LWS can be

used to derive the aggregated indices. Hence, the LWS approach and thead-hoc weights that are used to derive the CAI in this study both pro-vide robust and transparent results of coastal area classification.

3.3. Limitations and future prospects

The methodology applied in this work is based on GIS - MCDMapproach using weighted aggregation method. Detailed sensitivityanalysis was carried to demonstrate uncertainty associated with theweighting scheme and applied aggregation method. Still there are somelimitations applied to this work which are listed below and future di-rection of work that can address these limitations is suggested.

i. Applicability of CAI developed in this work is dependent on spatialresolution of available datasets in GIS. A better resolution of remotesensing imagery will give the better result under same testing con-ditions.

ii. The CAI index is heavily dependent on utility functions applied to

Fig. 5. Validation of CAI at selected location Colaba inSouth Mumbai and demonstration of the computationof the CAI for lower most left pixel (highlighted by thedark border). Membership function values are multi-plied by the layers weights (LW) and aggregated toobtain CAI.

R. Dhiman et al. Habitat International 71 (2018) 125–134

131

transform the GIS data into membership values used for aggrega-tion. In this work we have defined the utility function based on theliterature and our judgments, this has added the uncertainty to theresults. Different utility functions can be tested in future to deriveCAI.

iii. LWS method is applied for the aggregation of the weighted mem-bership values. It is well known that aggregation functions appliedchange the final index and hence other methods of multi-criteriadecision making should be applied and tested (Kalbar, Karmakar, &Asolekar, 2015).

iv. The test case for this work is Mumbai city and hence application ofCAI in other coastal areas should be carried out only after validationand calibration.

4. Conclusions

Coastal area classification presents numerous challenges that havebeen elaborated in the current work. The lack of scientific rational inIndian coastal zone classification policy has been identified in thiswork. A novel GIS-MCDM based CAI approach to classify urban coastalareas has been proposed to serve as scientific rational for coastal areaclassification.

The developed approach is transparent where physical character-istics of coastal areas are transformed into quantitative measurements.Utility based membership functions are developed and applied to eachof the coastal features. Weights are applied to each of the layers and

results are aggregated into one index, called CAI. The applicability ofthe CAI is demonstrated using Mumbai city case study. The robustnessof the results are validated using rigorous sensitivity analysis. The re-sults of sensitivity analysis showed that the weights and membershipfunctions used to derive CAI presented lower uncertainty than com-pared to the equal weights and random weight scenarios.

The methodology used is based on decision science where criteriarelies on physical utility of coastal features rather than the subjectiveinterest of stakeholders. This study provides the decision making fra-mework for location specific coastal area classification. Classification ofcoastal areas based on proposed CAI will facilitate planning based onscientific principles. CAI generates the distinct categories of coastalareas where most effective urban coastal management options can beproposed. The CAI computation method is modular and scalable innature and can be applied to different locations and datasets easily.Thus the CAI approach can further be used for developing integratedframework for sustainable coastal management.

The CAI results showed the usefulness to serve as decision supportapproach for urban planners and ULBs for classifying urban areas. Wehope that CAI will be taken up by the researchers and the practitionersand will be widely used for classifying urban coastal areas by policymakers, coastal managers, coastal communities and other associatedstakeholders.

Acknowledgements

Authors acknowledge detailed comments received from anonymousreviewers which helped us to improve the clarity of the manuscript.

Appendix A. Supplementary data

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.habitatint.2017.12.002.

References

Abadie, L. M., Galarraga, I., & de Murieta, E. S. (2017). Understanding risks in the light ofuncertainty: Low-probability, high-impact coastal events in cities. EnvironmentalResearch Letters, 12(1), 14017. http://dx.doi.org/10.1088/1748-9326/aa5254.

Fig. 6. (a) Comparison of CAI and relative mean of 1000 runs normalized with standard deviation for Ad-hoc weight scenario (z-score is shown on the axis). (b) Comparison of CAI andrelative mean of 1000 runs normalized with standard deviation for equal weight scenario (z-score is shown on the axis). (c) Comparison of CAI and relative mean of 1000 runs normalizedwith standard deviation for random weight scenario (z-score is shown on the axis).

Table 3Correlation matrix for membership function (pixel value) of different layers* (R2 values,Pearson Correlation).

LULC DEM GEO NDVI SLOPE SOIL

LULC 1 0.459 0.343 0.588 −0.006# 0.425DEM 1 0.393 0.267 0.136 0.443GEO 1 0.134 0.038 0.339NDVI 1 −0.025 0.276SLOPE 1 0.088SOIL 1

*All the values of R2 are significant at the level of 0.001 except the value marked with #.

R. Dhiman et al. Habitat International 71 (2018) 125–134

132

Alcántara-Ayala, I. (2002). Geomorphology, natural hazards, vulnerability and preven-tion of natural disasters in developing countries. Geomorphology, 47(2–4), 107–124.http://dx.doi.org/10.1016/S0169-555X(02)00083-1.

Almasri, M. N., & Kaluarachchi, J. J. (2005). Multi-criteria decision analysis for the op-timal management of nitrate contamination of aquifers. Journal of EnvironmentalManagement, 74(4), 365–381. http://dx.doi.org/10.1016/j.jenvman.2004.10.006.

Alves, F. L., Coelho, C., Coelho, C. D., & Pinto, P. (2011). Modelling coastal vulnerabilities- tool for decision support system at inter-municipality level. Journal of CoastalResearch, Suppl. Special Issue; Fort Lauderdale, SI(64), 966–970.

Ananda, J., & Herath, G. (2003a). Incorporating stakeholder v alues into regional forestplanning: A value function approach. Ecological Economics, 45, 75–90. http://dx.doi.org/10.1016/S0921-8009(03)00004-1.

Ananda, J., & Herath, G. (2003b). The use of Analytic Hierarchy Process to incorporatestakeholder preferences into regional forest planning. Forest policy and economics,5(1), 13–26.

Ananda, J., & Herath, G. (2009). Methods: A critical review of multi-criteria decisionmaking methods with special reference to forest management and planning.Ecological Economics, 68, 2535–2548.

Baby, S. (2013). AHP modeling for multicriteria decision-making and to optimise stra-tegies for protecting coastal landscape resources. International Journal of Innovation,Management and Technology, 4(2), 218–227. http://dx.doi.org/10.7763/IJIMT.2013.V4.395.

Baud, I., Scott, D., Pfeffer, K., Sydenstricker-Neto, J., & Denis, E. (2015). Reprint of:Digital and spatial knowledge management in urban governance: Emerging issues inIndia, Brazil, South Africa, and Peru. Habitat International, 46, 225–233. http://dx.doi.org/10.1016/j.habitatint.2015.01.018.

Benke, K. K., & Pelizaro, C. (2010). A spatial-statistical approach to the visualisation ofuncertainty in land suitability analysis. Journal of Spatial Science, 55(2), 257–272.http://dx.doi.org/10.1080/14498596.2010.521975.

Brail, R. K., & Klosterman, R. E. (2001). Planning support systems: Integrating geographicinformation systems, models, and visualization tools.

Brown, G., & Kyttä, M. (2014). Key issues and research priorities for public participationGIS (PPGIS): A synthesis based on empirical research. Applied Geography, 46,126–136. http://dx.doi.org/10.1016/j.apgeog.2013.11.004.

Bryan, M. (1988). Multi-criteria decision making (MCDM) techniques in planning.Progress in Planning, 30(part I), 1–84. https://doi.org/10.1016/0305-9006(88)90012-8.

Burford, M. A., Valdez, D., Curwen, G., Faggotter, S. J., Ward, D. P., & O'Brien, K. R.(2016). Inundation of saline supratidal mudflats provides an important source ofcarbon and nutrients in an aquatic system. Marine Ecology Progress Series, 545, 21–33.http://dx.doi.org/10.3354/meps11621.

Burrough, P., & McDonnell, R. (1998). Spatial information systems and geostatistics. In P.Burrough, & R. McDonnell (Eds.). Principles of geographical information systems.

Cenci, L., Disperati, L., Persichillo, M. G., Oliveira, E. R., Alves, F. L., & Phillips, M.(2017). Integrating remote sensing and GIS techniques for monitoring and modelingshoreline evolution to support coastal risk management. GIScience & Remote Sensing,1–21. http://dx.doi.org/10.1080/15481603.2017.1376370.

Cenci, L., Persichillo, M. G., Disperati, L., Oliveira, E. R., Alves, F. L., Pulvirenti, L., et al.(2015). Remote sensing for coastal risk reduction purposes: Optical and microwavedata fusion for shoreline evolution monitoring and modelling. International geoscienceand remote sensing symposium (IGARSS). Vol. 2015–Novem. International geoscience andremote sensing symposium (IGARSS) (pp. 1417–1420). IEEE. http://dx.doi.org/10.1109/IGARSS.2015.7326043.

Chang, H.-K., Liou, J.-C., & Chen, W.-W. (2012). Protection priority in the coastal en-vironment using a hybrid AHP-TOPSIS method on the Miaoli coast, Taiwan. Journalof Coastal Research, 280(2006), 369–374. http://dx.doi.org/10.2112/JCOASTRES-D-10-00092.1.

Chouhan, H., Parthasarathy, D., & Pattanaik, S. (2016). Urban development, environ-mental vulnerability and CRZ violations in India: Impacts on fishing communities andsustainability implications in Mumbai coast. Environment, Development andSustainability, 1–15. http://dx.doi.org/10.1007/s10668-016-9779-6.

Collins, M. G., Steiner, F. R., & Rushman, M. J. (2001). Land-use suitability analysis in theUnited States: Historical development and promising technological achievement.Environment Management, 5(28), 611–621.

Datta, S. N., Chakraborty, S. K., Jaiswar, A. K., & Ziauddin, G. (2010). A comparativestudy on intertidal faunal biodiversity of selected beaches of Mumbai coast. Journal ofEnvironmental Biology, 31(6), 981–986.

Drobne, S., & Lisec, A. (2009). Multi-attribute decision analysis in GIS: Weighted linearcombination and ordered weighted averaging. Informatica, 33, 459–474. http://dx.doi.org/10.1017/CBO9781107415324.004.

Eastman, J. R. (1995). GIS and decision making. Unitar.Feizizadeh, B., & Blaschke, T. (2014). An uncertainty and sensitivity analysis approach for

GIS-based multicriteria landslide susceptibility mapping. International Journal ofGeographical Information Science, 28(3), 610–638. http://dx.doi.org/10.1080/13658816.2013.869821.

Feng, L., Zhu, X., & Sun, X. (2014). Assessing coastal reclamation suitability based on afuzzy-AHP comprehensive evaluation framework: A case study of Lianyungang,China. Marine Pollution Bulletin, 89(1–2), 102–111. http://dx.doi.org/10.1016/j.marpolbul.2014.10.029.

Feizizadeh, B., & Blaschke, T. (2014). An uncertainty and sensitivity analysis approach forGIS-based multicriteria landslide susceptibility mapping. International Journal ofGeographical Information Science, 28(3), 610–638.

Fernandes, L., Ridgley, M. A., & Van’t Hof, T. (1999). Multiple criteria analysis integrateseconomic, ecological and social objectives for coral reef managers. Coral Reefs, 18(4),393–402. http://dx.doi.org/10.1007/s003380050217.

Gogate, N. G., Kalbar, P. P., & Raval, P. M. (2017). Assessment of stormwater

management options in urban contexts using Multiple Attribute Decision-Making.Journal of Cleaner Production, 142, 2046–2059. http://dx.doi.org/10.1016/j.jclepro.2016.11.079.

Gontier, M., Mörtberg, U., & Balfors, B. (2010). Comparing GIS-based habitat models forapplications in EIA and SEA. Environmental Impact Assessment Review, 30(1), 8–18.http://dx.doi.org/10.1016/j.eiar.2009.05.003.

Govindasamy, C., & Azariah, J. (1999). Seasonal variation of heavy metals in coastalwater of the Coromandel coast, Bay of Bengal, India. Indian Journal of Marine Sciences,28(3), 249–256.

Greening, L. A., & Bernow, S. (2004). Design of coordinated energy and environmentalpolicies: Use of multi-criteria decision-making. Energy Policy, 32(6), 721–735. http://dx.doi.org/10.1016/j.enpol.2003.08.017.

Hajkowicz, S. (2006). Multi-attributed environmental index construction. EcologicalEconomics, 57(1), 122–139. http://dx.doi.org/10.1016/j.ecolecon.2005.03.023.

Howard, A. F. (1991). A critical look at multiple criteria decision making techniques withreference to forestry applications. Canadian Journal of Forest Research, 21(11),1649–1659. http://dx.doi.org/10.1139/x91-228.

Janssen, R., & Rietveld, P. (1990). Multicriteria analysis and geographical informationsystems: An application to agricultural land use in The Netherlands. GeographicalInformation Systems for Urban and Regional Planning, 17, 129–139.

Joubert, A., Stewart, T. J., & Eberhard, R. (2003). Evaluation of water supply augmen-tation and water demand management options for the City of Cape Town. Journal ofMultiCriteria Decision Analysis, 12(1), 17–25. http://dx.doi.org/10.1002/mcda.342.

Kalbar, P. P., Birkved, M., Nygaard, S. E., & Hauschild, M. (2016). Weighting and ag-gregation in life cycle Assessment: Do present aggregated single scores provide cor-rect decision support? Journal of Industrial Ecology. http://dx.doi.org/10.1111/jiec.12520.

Kalbar, P. P., Karmakar, S., & Asolekar, S. R. (2015). Selection of wastewater treatmentalternative: Significance of choosing MADM method. Environmental Engineering andManagement Journal, 14(5), 1011–1020.

Keefer, D. L., Kirkwood, C. W., & Corner, J. L. (2004). Perspective on decision analysis.Decision Analysis, 1(1), 5–24. http://dx.doi.org/10.1287/deca.1030.0004.

Krishnamurthy, R. R., DasGupta, R., Chatterjee, R., & Shaw, R. (2014). Managing theindian coast in the face of disasters & climate change: A review and analysis of India'scoastal zone management policies. Journal of Coastal Conservation, 18(6), 657–672.http://dx.doi.org/10.1007/s11852-014-0339-7.

Le Cozannet, G., Garcin, M., Bulteau, T., Mirgon, C., Yates, M. L., Méndez, M., et al.(2013). An AHP-derived method for mapping the physical vulnerability of coastalareas at regional scales. Natural Hazards and Earth System Science, 13(5), 1209–1227.http://dx.doi.org/10.5194/nhess-13-1209-2013.

Longley, P., Goodchild, M. F., Maguire, D., & Rhind, D. (2005). Geographical informationsystems and science. John Wiley & Sons339–340.

Malczewsk, J. (1999). GIS and multicriteria decision analysis.Malczewski, J. (2004). GIS-based land-use suitability analysis: A critical overview.

Progress in Planning, 62(1), 3–65. http://dx.doi.org/10.1016/j.progress.2003.09.002.Malczewski, J. (2006). GIS-based multicriteria decision analysis: A survey of the litera-

ture. International Journal of Geographical Information Science, 20(December),703–726. http://dx.doi.org/10.1080/13658810600661508.

Mani Murali, R., Ankita, M., Amrita, S., & Vethamony, P. (2013). Coastal vulnerabilityassessment of Puducherry coast, India, using the analytical hierarchical process.Natural Hazards and Earth System Sciences, 13(12), 3291–3311. http://dx.doi.org/10.5194/nhess-13-3291-2013.

Mendoza, G. A., & Martins, H. (2006). Multi-criteria decision analysis in natural resourcemanagement: A critical review of methods and new modelling paradigms. ForestEcology and Management, 230(1–3), 1–22. http://dx.doi.org/10.1016/j.foreco.2006.03.023.

Munda, G., Nijkamp, P., & Rietveld, P. (1993). Qualitative multicriteria evaluation forenvironmental management. Ecological Economics, 10(2), 97–112. https://doi.org/10.1016/0921-8009(94)90002-7.

Murali, R. M., Dhiman, R., Choudhary, R., Seelam, J. K., Ilangovan, D., & Vethamony, P.(2015). Decadal shoreline assessment using remote sensing along the central Odishacoast, India. Environmental Earth Sciences, 74(10), 7201–7213. http://dx.doi.org/10.1007/s12665-015-4698-7.

Murthy, R. C., Rao, Y., & Inamdar, A. B. (2001). Integrated coastal management ofMumbai metropolitan region. Ocean & Coastal Management, 44(5–6), 355–369.http://dx.doi.org/10.1016/S0964-5691(01)00055-2.

Nair, P., & Gopinathan, C. (1981). Productivity of the exclusive economic zone of India.Journal of the Marine Biological Association of India, 23(1&2), 48–54.

Nayak, B. U., Chandramohan, P., & Desai, B. N. (1992). Planning and management of thecoastal zone in India - a perspective. Coastal Management, 20(4), 365–375. http://dx.doi.org/10.1080/08920759209362185.

Newton, A., Carruthers, T. J. B., & Icely, J. (2012). The coastal syndromes and hotspots onthe coast. Estuarine, Coastal and Shelf Science, 96(1), 39–47. http://dx.doi.org/10.1016/j.ecss.2011.07.012.

Nijkamp, P., Rietveld, P., & Voogd, H. (1990). Multicriteria evaluation in physical planning.Orencio, P. M., & Fujii, M. (2013). A localized disaster-resilience index to assess coastal

communities based on an analytic hierarchy process (AHP). International Journal ofDisaster Risk Reduction, 3(1), 62–75. http://dx.doi.org/10.1016/j.ijdrr.2012.11.006.

Pacione, M. (2006). Mumbai. Cities, 23(3), 229–238. http://dx.doi.org/10.1016/j.cities.2005.11.003.

Parthasarathy, D. (2011). Hunters, gatherers and foragers in a Metropolis: Commonisingthe private and public in Mumbai. Economic & Political Weekly, XLVI(50), 54–63.

Pourebrahim, S., Hadipour, M., Mokhtar, M. B., & Taghavi, S. (2014). Application ofVIKOR and fuzzy AHP for conservation priority assessment in coastal areas: Case ofKhuzestan district, Iran. Ocean and Coastal Management, 98, 20–26. http://dx.doi.org/10.1016/j.ocecoaman.2014.05.009.

R. Dhiman et al. Habitat International 71 (2018) 125–134

133

Ramachandran, S. (1999). Coastal zone management in India—problems, practice andrequirements. Perspectives on integrated coastal zone management.

Ramesh, R., Lakshmi, A., George, A., & Purvaja, R. (2015). Building resilient coastalcommunities by enabling participatory action: A case study from India. In B.Glavovic, M. Kelly, R. Kay, & T. Ailbhe (Eds.). Climate change and the coast (pp. 169–184). CRC Press. http://dx.doi.org/10.1201/b18053-11.

Renwick, S. (2015). Prioritisation of food security by decision makers in the caribbean, astudy of three caribbean territories: Trinidad and Tobago, Belize and Barbados.Proceedings of the 30th West Indies agricultural economics conference 30th June – 6 Thjuly, 2013, port of Spain, Trinidad.

Rog, S. M., Clarke, R. H., & Cook, C. N. (2017). More than marine: Revealing the criticalimportance of mangrove ecosystems for terrestrial vertebrates. Diversity andDistributions, 23(2), 221–230. http://dx.doi.org/10.1111/ddi.12514.

Schmoldt, D., Kangas, J., Mendoza, G. A., & Pesonen, M. (2001). The analytic hierarchyprocess in natural resource and environmental decision making. Journal of ChemicalInformation and Modeling, 3. http://dx.doi.org/10.1007/978-94-015-9799-9.

Shirodkar, P. V., Mesquita, A., Pradhan, U. K., Verlekar, X. N., Babu, M. T., & Vethamony,P. (2009). Factors controlling physico-chemical characteristics in the coastal watersoff Mangalore-A multivariate approach. Environmental Research, 109(3), 245–257.http://dx.doi.org/10.1016/j.envres.2008.11.011.

Smith, K., & Petley, D. N. (2008). Environmental Hazards: Assessing risk and reducing dis-aster. Routledge| Taylor & Francis Group.

Sonak, S., Pangam, P., & Giriyan, A. (2008). Green reconstruction of the tsunami-affectedareas in India using the integrated coastal zone management concept. Journal ofEnvironmental Management, 89(1), 14–23. http://dx.doi.org/10.1016/j.jenvman.2007.01.052.

Srinivas, H., & Nakagawa, Y. (2008). Environmental implications for disaster prepared-ness: Lessons learnt from the Indian Ocean tsunami. Journal of EnvironmentalManagement, 89(1), 4–13. http://dx.doi.org/10.1016/j.jenvman.2007.01.054.

Sudha Rani, N. N. V., Satyanarayana, A. N. V., & Bhaskaran, P. (2015). Coastal vulner-ability assessment studies over India: A review. Natural Hazards, 77(1), http://dx.doi.org/10.1007/s11069-015-1597-x.

Surjan, A., Parvin, G. A., Atta-ur-Rahman, & Shaw, R. (2016). Expanding coastal Cities:An increasing risk. Urban disasters and resilience in Asia (pp. 79–90). . https://doi.org/10.1016/B978-0-12-802169-9.00006-9.

Vaz, E. (2014). Managing urban coastal areas through landscape metrics: An assessmentof Mumbai's mangrove system. Ocean & Coastal Management, 98, 27–37. http://dx.doi.org/10.1016/j.ocecoaman.2014.05.020.

Yeh, A. G., & Li, X. (1998). Sustainable land development model for rapid growth areasusing GIS. International Journal of Geographical Information Science, 12(2), 169–189.http://dx.doi.org/10.1080/136588198241941.

Zhou, P., Ang, B., & Poh, K. (2006). Decision analysis in energy and environmentalmodeling: An update. Energy, 31(14), 2604–2622. http://dx.doi.org/10.1016/j.energy.2005.10.023.

R. Dhiman et al. Habitat International 71 (2018) 125–134

134